Published in Vadose Zone Journal 3:901-908 (2004)
© 2004 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
SPECIAL SECTION: RESEARCH ADVANCES IN VADOSE ZONE HYDROLOGY THROUGH SIMULATIONS WITH THE TOUGH CODES
Application of Nonisothermal Multiphase Modeling to In Situ Soil Remediation in Söderkulla
Terhi Klinga,*,
Juhani Korkealaaksoa and
Jukka Saarenpääb
a Technical Research Centre of Finland, P.O. Box 1804, FIN-02044 VTT, Finland
b Finnish Road Enterprise, P.O. Box 382, FIN-33101 Tampere, Finland
* Corresponding author (terhi.kling{at}vtt.fi)
Received 28 August 2004.
 |
ABSTRACT
|
|---|
Nonisothermal multiphase modeling with the T2VOC numerical simulator was used as an integration tool in an in situ remediation project in which trichloroethylene (TCE)-contaminated soil was treated with thermally (steam) enhanced soil vapor extraction. Numerical simulation studies were employed in the iterative conceptualization of the migration of the contaminants and the effects of different treatment operations, in the planning of optimal system design and control of the remediation processes, as well as for integration of information obtained from field investigations during different phases of the project. Modeling proved to be a valuable tool in planning and decision making. It was, however, difficult to forecast the costs or the duration of the project, since the actual TCE mass and distribution was poorly known. The power of modeling lies in its ability to compare the effect of different strategies or measures on the cleanup efficiency, and in studying the associated effects during the treatments.
Abbreviations: DNAPL, dense nonaqueous phase liquid • NAPL, nonaqueous phase liquid • TCE, trichloroethylene • VOC, volatile organic compound
|
INTRODUCTION
|
|---|
SOIL AT THE SöDERKULLA ESTATE in Sibbo Municipality, Southern Finland, is contaminated with TCE originating from a metal workshop that operated in the area more than 30 yr ago. The soil consists of a 3- to 4-m-thick fine-textured (clay) layer surrounded by permeable sand. The contaminants are mainly located in the lower parts of the clay layer above the water table. Trichloroethylene concentrations in soil of >10 mg kg–1 were found in a circular area with a diameter of 15 to 20 m, with the highest concentration 231 mg kg–1. Based on interpolation, the total amount of TCE in this area was estimated to be about 110 kg. At the time of the soil survey the highest detected TCE concentrations in the underlying aquifer were about 700 µg L–1. A site plan is shown in Fig. 1
.
View larger version (39K):
[in this window]
[in a new window]
|
Fig. 1. Map showing the most contaminated area, where trichloroethylene (TCE) concentrations in the soil were >10 mg kg–1. Lines C and E are tomography lines (Fig. 3).
|
|
TCE is a dense nonaqueous phase liquid (DNAPL); that is, it is denser than water and forms a separate nonaqueous phase liquid (NAPL) phase in the soil. Although NAPL at saturations less than residual saturation values does not migrate, it can serve as an important source of secondary contamination by partitioning into the soil gas phase through volatilization, and into soil water through dissolution. The TCE contamination can thus cause ongoing groundwater contamination if it is not remediated.
Because the area is a historical site and a protected building (an old cowshed) was located partly on top of the most contaminated area, remediation by excavation was not judged to be cost effective. In situ treatment was therefore selected. A remedial system was designed (Fig. 2)
in which air and steam are injected under the clay layer to thermally enhance the cleanup by soil vapor extraction. Contaminant vapors are treated by catalytic oxidation. Groundwater is pumped and treated by means of an activated carbon filter. Soil vapor extraction is a widely used technology for the removal of chlorinated solvents from the vadose zone (Relander et al., 1999). When the contaminants are located in low permeability areas, soil vapor extraction with air injection only is relatively ineffective. The air then flows preferentially through highly permeable regions, thus making it difficult to remove even highly volatile NAPL from the low permeability regions. The transport of contaminants out of the low permeability regions occurs mainly by diffusion. Remediation can be enhanced by thermal methods to provide high gradients of NAPL–vapor concentration in the gas phase (Class and Helmig, 2002).
Steam injection has proved to be particularly effective due to the high energy injection rate and the ability of steam to displace pore fluids, even in saturated soils (Udell, 1996). There are several mechanisms acting on the contaminants during steam injection: physical displacement, a decrease in the capillary and interfacial forces between the fluids and the porous media, enhanced desorption of contaminants from the porous solids, a decrease in the viscosity of the organic phase, steam distillation, and steam displacement (Davis, 1998).
Steam injection should be used with air coinjection since injection of pure steam could cause a NAPL to condense at the temperature front, thereby making it mobile and allowing downward flow, which in the case of a DNAPL could cause the contaminants to enter deeper zones of the aquifer. Using steam–air co-injection makes it possible to control the condensation so mobilization of the condensate is prevented. The noncondensible air continuously transports a part of the NAPL vapor to the extraction wells (Class and Helmig, 2002).
One way to raise the cost effectiveness of steam injection may be cyclic injection. Halting steam injection but continuing vapor extraction will cause depressurization of the steam zone. To bring the temperature and pressure back into thermodynamic equilibrium, the temperature will be reduced by evaporation of the residual water and contaminants, which are then removed from the system by vacuum extraction (Davis 1998). Itamura and Udell (1995) showed that cyclic steam injection with continuous vacuum extraction will always reduce the amount of steam required to meet a given cleanup level, and may also reduce the overall cleanup time.
We present simulation results of TCE remediation activities at the Söderkulla site. The simulations demonstrate how modeling can be used to support and improve real-world remediation projects.
|
MODELING
|
|---|
Simulations were performed with T2VOC (Falta et al., 1995), a numerical simulator for three-phase, three-component, nonisothermal flow of water, air, and a volatile organic compound (VOC) in multidimensional heterogeneous porous media. T2VOC is an extension of the TOUGH2 (Pruess et al., 1999) general-purpose simulation program, which uses a general integral finite difference formulation of the multiphase, multicomponent mass and energy balance equations.
In the T2VOC formulation the three fluid components (air, water, and a volatile, water-soluble organic chemical) may be present in different proportions in any of the three phases (gas, aqueous, and NAPL), except that the (usually low) solubility of water in the NAPL phase is neglected. Each phase flows in response to pressure and gravitational forces according to the multiphase extension of Darcy's Law, including the effects of relative permeability and capillary pressure between the phases. Transport of the three mass components also occurs by multicomponent diffusion in the gas phase, without consideration of molecular diffusion in the aqueous and NAPL phases, or by hydrodynamic dispersion.
The three phases are assumed to be in local chemical and thermal equilibrium. Heat transfer occurs due to conduction, multiphase convection, and gaseous diffusion. The heat transfer effects of phase transitions between the NAPL, aqueous, and gas phases are fully accounted for by considering the transport of both latent and sensible heat. The porous media thermal conductivity depends on the phase saturations and the chemical characteristics of the NAPL. Fluid properties are generally dependent on temperature and pressure. The calculations presented below do not include adsorption or decay of VOC through biodegradation, although the code provides the possibility of taking these into account.
|
PRELIMINARY SIMULATIONS
|
|---|
No data were available about either the original TCE spill or the exact time of the spill. In the characterization of the contamination in the area and in the preliminary conceptualization process, the soil and groundwater data were combined with the information from a number of simulations. The soil parameters (Tables 1–3) used in the simulations were selected on the basis of the literature (Mercer and Cohen, 1990; Pruess, 1992; Sundberg, 1988; Jääskeläinen et al., 1982) and considering Söderkulla site conditions. Annual precipitation was taken into account by assuming an infiltration rate of 300 mm yr–1.
If a spill of TCE occurs near the soil surface, the contaminant percolates downward through the unsaturated zone toward the water table. In the unsaturated zone, water present on the solid grains is the wetting phase, while the NAPL is the wetting phase with respect to soil gas. The NAPL does not displace water from the grain surfaces, but does displace water that is not strongly held, as well as the soil gas. Due to capillary forces and varying hydraulic conductivity, the plume also exhibits a tendency to spread horizontally. Even a relatively thin, low permeability layer will inhibit downward percolation and force the contaminant to spread laterally. If the layer is discontinuous, the NAPL will eventually spill over and continue to move downward. In the capillary fringe, where water saturates a large proportion of the pores, the relative permeability to the immiscible liquid declines; consequently, there is a tendency for NAPL to spread (Domenico and Schwartz, 1990).
In Söderkulla the main geological feature influencing the spread of NAPL is a 3- to 4-m thick clay layer surrounded by permeable sand. Electrical resistivity tomography (Fig. 3)
results combined with the soil sampling data revealed that the clay layer is not continuous. Depending on the location of the spill, many possible preferential flow paths may exist along the top of the clay, through the clay or into the clay.
Spreading of TCE at the heterogeneous site with a clay layer was simulated using a vertical two-dimensional T2VOC-model. In the simulations (Fig. 4)
, NAPL ponds on top of the clay, spreads laterally, spills over, and then continues to flow downwards. The NAPL subsequently spreads in the capillary fringe, continues through the saturated zone, and eventually ponds at the bottom of the model. No NAPL penetrated into the clay when the water content (w) of the clay was about 35%. (In Söderkulla the measured water content of the clay in the soil samples was 33-44%.) Our simulations showed that only in cases of a long period of drying just before the spill would NAPL penetrate partly into the clay layer.
View larger version (17K):
[in this window]
[in a new window]
|
Fig. 4. Simulated two-dimensional spreading of the NAPL phase in a vertical cross section during a hypothetical spill of trichloroethylene (TCE). The color contours in the figure represent NAPL saturation.
|
|
When the simulation was continued for 30 yr after the spill, soil concentrations gradually reduced by volatilization, dissolution, and solute transport. The highest concentrations occurred at first in the high permeability areas, while at the end of the simulation the contamination was mainly confined to the low permeability areas. According to soil survey data from Söderkulla the maximum concentrations at the present time are in the lower parts of the clay layer (at depths of 3.5 to 4 m) and in the basal till above the bedrock (at depths of 14.3 to 14.5 m). Only small amounts of TCE remain in the thick sand layer between these deposits.
A zero-dimensional batch model was used to study the distribution of TCE between NAPL, aqueous, and gaseous phases for different temperatures and soil conditions (Table 4). The highest observed soil TCE concentration in Söderkulla was 231 mg kg–1 in the clay (above the ground water table), and 11.1 mg kg–1 in the sand (below the water table). These relatively low soil concentrations indicate that TCE was predominantly confined to the aqueous and gas phases. The fact that the observed TCE concentrations were below the solubility limit does not necessarily imply the absence of a NAPL phase; we believe that some NAPL still was present in the clay layer near the most contaminated sampling points.
|
SIMULATIONS OF THE PILOT PHASE
|
|---|
During the pilot phase (22 Feb.–31 May 2001), the remediation technique was tested under field conditions. In the first stage air was injected (120 kg h–1) under the clay layer from two vertical wells (I6 and I7 in Fig. 5)
in the center, and soil vapor was extracted from four vertical wells (E8, E9, E10, and E11 in Fig. 5) located around the injection wells and from five horizontal wells (H1, H2, H3, H4, and H5 in Fig. 5) above the clay. After 1 mo, steam was injected (300 kg h–1, 200°C) into the middle wells (I6 and I7). After some time, the steam penetrated into the clay and the injection had to be stopped. Fracturing of the clay may have helped to clean up the highly contaminated zones in the middle of the area, but the wells were damaged and could no longer be used for steam injection. The breakthrough took place because the steam injection rate was too high, especially with air co-injection, and because the filter section of the well was not entirely located in the high permeability zone. The situation was handled by changing to a smaller steam generator (120 kg h–1, 120°C) and drilling new wells (I12 and I13 in Fig. 5) for pure steam injection.
View larger version (21K):
[in this window]
[in a new window]
|
Fig. 5. The simulation model used for the pilot scale simulations. I6, I7, I12 and I13 are injection wells; E8-E11 (vertical) and H1-H5 (horizontal) are extraction wells.
|
|
The subsurface was intensively monitored during steam injection to collect data for model calibration and for designing a full-scale treatment system. Temperature was measured at 0.5-m intervals along nine vertical lines to a depth of 8 m. The maximum temperature in the soil was 102.8°C at a depth of 4.5 to 5.5 m. The maximum temperature at a lateral distance of 5 m from the injection wells was 82.3°C.
The pilot phase remedial actions were simulated using the three-dimensional T2VOC-model shown in Fig. 9. Throughout the pilot project all new data obtained from the field were used to make the model more closely resemble the real situation. Layer thicknesses and the groundwater levels, as well as the initial and boundary conditions, were modified by "trisl" and "error" until good agreement was obtained with the field data.
View larger version (54K):
[in this window]
[in a new window]
|
Fig. 9. Planned full-scale remediation system and simulated temperature distribution along the lower edge of the clay layer after 2 wk of cyclical-rotational steaming from Wells I12 through I15. In these simulations the former Injection Well I7 was used for vapor extraction.
|
|
The saturated zone was discretized into three layers, and the water table was set at the level measured in standpipes. The upper boundary was assumed to be closed because the site was covered with an airtight cover (this assumption was later changed to allow heat flow by conduction through the boundary). The initial soil conditions were calculated using simulations with an infiltration of 300 mm yr–1 (half of the annual precipitation). With these assumptions and the parameter values presented in Tables 1 through 3, the simulated soil moisture content was in good agreement with the soil moisture content distribution (Fig. 6)
measured using a radiometric in situ method.
View larger version (16K):
[in this window]
[in a new window]
|
Fig. 6. Simulated and measured soil moisture contents w (clay at a depth of 1–4 m, groundwater level at a depth of 5 m.)
|
|
A constant temperature of 7°C, corresponding to the average yearly temperature in Southern Finland, was specified along the outer model boundaries, which were outside the range of influence of the wells. The final calibration was made against temperature data. Fracturing of the clay was simulated using high-permeability elements (k = 10–11 m2).
In the simulations several approaches were used to describe the contamination process. To obtain information about the potential cleanup effectiveness in the area, a uniformly distributed initial contamination was used in the entire area. In a more realistic approach, the initial contaminant distribution was interpolated from the soil sample concentration data. In an intermediate approach, the target area was divided into two parts: a slightly contaminated area (18 by 18 by 5 m) with a soil concentration of 25 mg kg–1, and a highly contaminated area (10 by 10 by 3 m) in the middle with a soil concentration of 225 mg kg–1. The total initial amount of TCE in each case was 156 kg.
Simulation results are presented in Fig. 7 and 8
. Notice that the implemented cleanup operations are limited mostly to the middle and the northern sides of the area, while the southern part of the area is left almost unaffected.
View larger version (54K):
[in this window]
[in a new window]
|
Fig. 7. Simulation results showing steam break through the clay layer around Injection Wells I6 and I7. The gray area (around Wells I6 and I7) has temperatures exceeding 60°C. The color contours represent trichloroethylene (TCE) concentrations in the water.
|
|
View larger version (55K):
[in this window]
[in a new window]
|
Fig. 8. Simulation results showing the effects of the newly installed Injection Wells I12 and I13. The gray area (around Well I13) has temperatures exceeding 60°C. The color contours represent trichloroethylene (TCE) concentrations in the water.
|
|
The simulated amount of TCE extracted during the cleanup was compared with measurements based on the temperature difference in the catalytic oxidation process. During the first 2 mo the agreement was good: in the simulations 44 kg of TCE was removed, while (according to the measurements) the amount of TCE extracted was 43 kg. Since the effect of cooling on the catalytic oxidation process was not taken into account, the actual amount of TCE removed should be larger. On the other hand, it was not possible to consider in the measured extraction rate all components that were oxidized during the steam injection process. This would suggest that the actual value is somewhat smaller than the measured one. In the full-scale project, the monitored extraction rate based on catalytic oxidation was compared with the destruction rate calculated from vapor samples.
At the end of the pilot project the total amount of TCE extracted was estimated to be 100 to 150 kg, which was more than the original estimate of the total amount of contaminant in the soil. This extraction amount could only be achieved in the model if the initial TCE inventory was increased to 300 kg. This finding illustrates the difficulty in predicting the remediation efficiency. In the case of a NAPL chemical, the contaminant distribution in the subsurface tends to be highly heterogeneous, which makes it difficult to determine the total amount of contaminant initially in the soil. If the initial contaminant distribution is unknown, it is impossible to accurately predict cleanup times. The power of the modeling lies in its ability to compare the effect of different strategies or measures on the cleanup efficiency, or in studying associated effects or risks during treatment.
Drying of the clay is one risk that should receive attention because of possible soil compression. The drying effect was numerically studied by varying the injection–extraction combinations. Simulation results showed that with air injection, drying occurred throughout the soil matrix between the wells, depending on the dryness of the injected air. With totally dry air, the soil moisture content of the clay decreased after 8 mo by 13% in the 1-m-thick bottom layer, and by 7% in the next layer. With air–steam co-injection, some drying occurred when the temperatures were at a maximum. However, after steaming, the soil moisture content soon reverted to the original value because of the effects of steam condensation. Because of the possible risks of clay drying, the foundation of the cowshed was monitored for settling using radiometric in situ measurement of soil moisture content to assure that no damage would occur.
|
DESIGNING A FULL-SCALE REMEDIATION SYSTEM
|
|---|
The calibrated model was used in designing a full-scale treatment system capable of heating the entire contaminated area. The planned full-scale system and the simulated temperature distribution along the lower edge of the clay after 2 wk of steaming are presented in Fig. 9
. Two new injection wells (I14 and I15) were needed to heat the whole area, while two new vertical extraction wells (E16 and E17) were required for effective collection of the contaminant vapors. The water table had to be lowered by 0.5 to 1 m. The new installations were implemented in November 2001 according to this design.
During the full-scale operations two steam generators were used for steam injection: a small generator (20 kg h–1) used for steam–air co-injection, and a larger generator (360 kg h–1) for cyclical steam injection distributed over two to four injection wells at a time. Different injection–extraction combinations, cycles, and rotations were compared in the simulations to find the most cost-effective and time-saving strategies (Fig. 10) . The simulations showed that the most time-saving strategies were those in which steam injection was made rotationally from Wells I12, I13, I14, and I15 while maintaining air injection all the time in the middle (I7) by means of a small steam co-injection. The more frequently the steam injection locations were changed, the more efficient the cleanup process.
View larger version (13K):
[in this window]
[in a new window]
|
Fig. 10. Proportion of trichloroethylene (TCE) (%) extracted with in the simulations by cyclical steam injection (S1) and rotational steam injection (S2).
|
|
It was not possible for field operations to always rigorously follow the planned steaming strategies. Some delays occurred in getting the system to work properly, and with the high levels of steam injection somebody had to be present at the site to monitor and control the steam generator. This meant that active operations (changing injection wells, using a high steam injection rate) were limited to daytime working hours. We believe that in the future these operations can be partly automated and that the operational scheme (defining pumping volumes and time schedules, steam temperatures, and injection locations) can be further optimized by means of optimization techniques (Finsterle, 2000) to reduce the remediation costs.
|
DESIGN CONSIDERATIONS
|
|---|
Simplified simulations were made with the calibrated model to compare different design options. In these simulations, the contamination distribution was interpolated from the soil sampling data, while some extra TCE was added near the most contaminated area so the total amount of contaminant in the model was 152.6 kg. A medium-sized steam generator (120 kg h–1) and air co-injection (120 kg h–1) were used as the base case.
When using steam injection it should be noted that steam flows through the high permeability areas and that heating of the low permeability areas occurs through heat conduction. This makes the heating a slow process in the case of a thick layer of low permeability, such as at the Söderkulla site. Steam injection, however, makes the remediation process much faster than pure air injection, which can be seen in the simulation results shown in Fig. 11
.
One of the most important design considerations is the steam injection rate. When injection pressures greater than the overburden pressure are used, fracturing can occur in the overburden, thereby allowing steam to escape to the surface (Davis, 1998). Simulations were performed to compare the efficiency of the original design without fracturing with a situation where fracturing takes place but steam injection is continued in the same well. In this simulation, steam and air were injected into the central wells (I6 and I7), while fracturing was described by means of high permeability elements (k = 10–11 m2). The results (Fig. 12)
show that, in the beginning, fracturing made the cleanup process much more efficient, but that, because the effects were focused on a smaller area, the system soon became ineffective compared with the case where no fracturing took place. One way to avoid fracturing is to divide the injection among more wells, thus reducing the local pressures.
The system could be further optimized if the steam–air co-injection rate could be adjusted more freely. Simulations were made to compare steam–air injections with different steam or air injection rates. In the simulations the efficiency was more sensitive to the steam injection rate than to the air injection rate. This is a direct consequence of the temperature behavior being most sensitive to the steam injection rate (Class, 2001) and the cleanup being driven by the temperature distribution in the contaminated low permeability zones (Class and Helmig, 2002).
|
SUMMARY AND CONCLUSIONS
|
|---|
Nonisothermal multiphase modeling was used as an integration tool in the planning, control, and optimization of in situ treatment of TCE contaminated soil in Söderkulla. For the purposes of characterizing the contamination in the area and developing a preliminary conceptual model, soil and groundwater data were combined with information from a number of simulations. Modeling was also applied in the planning of optimal system design and control of the processes during remediation, as well as for integration of information obtained from the field investigations during the different phases of the project.
The main steps in the modeling were:
- Vertical two-dimensional models were used for the qualitative transport simulations.
- Batch models were used to study the distribution of TCE between the phases for different temperatures and soil conditions.
- Three-dimensional site-specific models were used for qualitative and quantitative remediation simulations.
- The model was calibrated against field data collected during the pilot phase.
- The calibrated model was used in planning a full-scale treatment system.
- Simplified simulations were made to compare different design options.
- During the full-scale project the model was used to compare different remediation strategies.
The main difficulties in the modeling were:
- Estimating the initial contaminant distributions—if the initial contaminant distribution or mass in the soil is unknown, then it is impossible to predict the cleanup time.
- Estimating the effect of steam on the permeability of the soil—steam can significantly alter the permeability structure of a soil, making it necessary to assure the validity of the model with field measurements and to be prepared to make further calibrations when needed.
Modeling can be used as a decision-making tool to help differentiate the pros and cons of competing remediation alternatives. The power of modeling lies in its ability to compare the effect of different strategies or measures on the cleanup efficiency, or in studying associated effects during the treatments. Modeling also makes it possible to handle the interconnected effects of the processes, and thus minimize potential risks associated with the project.
|
APPENDIX
|
|---|
CSoil concentration (mg kg–1)tTime (wk, mo)TTemperature (°C)kSoil permeability (m2)nSoil porositySLSoil liquid saturationwSoil moisture content (%)DrockRock grain density (kg m–3)SPHTRock grain specific heat (J kg–1 °C)CwetFormation heat conductivity under fully liquid-saturated conditions (W m–1 °C)CdryFormation heat conductivity under desaturated conditions (W m–1 °C)SwrResidual water saturationSnrResidual NAPL saturationSgrResidual gas saturationnStoneConstant used in the Stone's relative permeability function (Stone, 1970)SmConstant used in the Parker's capillary pressure function (Parker et al., 1987) SRC="/math/agr.gif" ALT="{alpha}" BORDER="0">gnConstant used in the calculation of gas–NAPL capillary pressure (Parker et al. 1987)
nwConstant used in the calculation of NAPL–water capillary pressure (Parker, 1987)X,Y,ZCartesian coordinates (m)
|
ACKNOWLEDGMENTS
|
|---|
The remediation project was carried out cooperatively by the Technical Research Centre of Finland (VTT), the Finnish Road Enterprise (FRE), and MB Envirotech AB (Sweden). We would like to thank Mika Kaakkomäki (FRE) and Jonny Bergman (MB Envirotech AB) for implementing the field experiments. We are especially grateful to Ilona Nokela, the technical manager of Sibbo municipality, for her support. Helpful discussions with Kent Udell, Karsten Pruess, and Chris Doughty are appreciated. For a review of the manuscript and the suggestion of improvements we thank Stefan Finsterle, Curtis Oldenburg and Rien van Genuchten. Comments by three anonymous referees are gratefully acknowledged.
|
REFERENCES
|
|---|
- Class, H. 2001. Theorie und numerische Modellierung nichtisothermer Mehrphasenprozesse in NAPL-kontaminierten porösen Medien. Ph.D. Thesis, Heft 105. Institut für Wasserbau, Universität Stuttgart, Germany.
- Class, H., and R. Helmig. 2002. Numerical simulation of non-isothermal multiphase multicomponent processes in porous media. 2. Applications for the injection of steam and air. Advances in Water Resources 25:551–564.
- Davis, E.L. 1998. Steam injection for soil and aquifer remediation. EPA Ground Water Issue, EPA/540/S-97/505. USEPA, Washington, DC.
- Domenico, P.A., and F.W. Schwartz. 1990. Features of nonaqueous contaminant spreading. p. 601–604. In Physical and chemical hydrogeology. John Wiley & Sons, New York.
- Falta, R.W., K. Pruess, S. Finsterle, and A. Battistelli. 1995. T2VOC user's guide. Report LBL-36400. Lawrence Berkeley Natl. Lab., Berkeley, CA.
- Finsterle, S. 2000. Demonstration of optimization techniques for groundwater plume remediation. Rep. LBNL-46746. Lawrence Berkeley Natl. Lab., Berkeley, CA.
- Itamura, M.T., and K.S. Udell. 1995. An analysis of optimal cycling time and ultimate chlorinated hydrocarbon removal from heterogeneous media using cyclic steam injection. p. 651–660. In Proc. ASME Heat Transfer and Fluids Engineering Divisions HTD. Vol. 321. ASME, New York.
- Jääskeläinen, R., M. Rantamäki, and M. Tammirinne. 1982. Geotechnics (In Finnish.) Otapaino, Espoo, Finland.
- Mercer, J.W., and R. M. Cohen. 1990. A review of immiscible fluids in the subsurface: Properties, models, characterization and remediation. J. Contam. Hydrol. 6:107–163.
- Parker, J.C., R.J. Lenhard, and T. Kuppusamy. 1987. A parametric model for constitutive properties governing multiphase flow in porous media. Water Resour. Res. 23:618–624.
- Pruess, K. 1992. Analysis of flow processes during TCE infiltration in heterogeneous soils at the Savannah River Site, Aiken, South Carolina. Rep. LBL-29400. Lawrence Berkeley Natl. Lab., Berkeley, CA.
- Pruess, K., C. Oldenburg, and G. Moridis. 1999. TOUGH2 user's guide. Version 2.0. Rep. LBNL-43134. Lawrence Berkeley Natl. Lab., Berkeley, CA.
- Relander, A., J. Aho, and J. Korkealaakso. 1999. Review on in situ remediation technologies for contaminated soil and groundwater. (In Finnish.) VTT–Tiedotteita–Meddelanden–Research Notes. Technical Research Centre of Finland, Espoo, Finland.
- Stone, H.L. 1970. Probability model for estimating three-phase relative permeability. Trans. SPE AIME 249:214–218.
- Sundberg, J. 1988. Thermal properties of soils and rocks. Rep. 35. Swedish Geotechnical Inst., Linköping, Sweden.
- Udell, K.S. 1996. Heat and mass transfer in clean up of underground toxic wastes. p. 330–405. In C.-L.Tien (ed.) Annual review of heat transfer. Begell House, New York.
TOP
ABSTRACT
INTRODUCTION
